In 1972 Congress, intending “to restore and maintain the chemical, physical, and biological integrity of the Nation’s waters,” amended the Federal Water Pollution Control Act (Act), 33 U.S.C. § 1251 'et seq. Congress established national pollution goals to be achieved by specific dates. By July 1, 1977, industries discharging pollutants into the nation’s waters were to have achieved “the best practicable control technology currently available (BPT).” Section 301(b)(1)(A), 33 U.S.C. § 1311 (b)(1)(A). By 1983, industry is to achieve “the best available technology economically achievable (BEA).” Section 301(b)(2)(A), 33 U.S.C. § 1311 (b)(2)(A). The Environmental Protection Agency (EPA or Agency) was entrusted with the responsibility of defining and policing the efficient and prompt achievement of these goals. Section 304, 33 U.S.C. § 1314 .

This case involves a challenge to regulations promulgated by the Agency establishing effluent guidelines for the Canned and Preserved Seafood Processing Point Source Category. 40 C.F.R. §§ 408.10 et seq. See also 40 Fed.Reg. 55770 et seq. (Dec. 1,1975). Petitioner Association of Pacific Fisheries is a trade association representing canners and fresh and frozen fish processors in affected subcategories. The remaining petitioners process seafood in all the subcategories at issue on this appeal.

The EPA’s regulations governing the fish processing industry were promulgated in *802 two phases. 1 In the first phase, the EPA issued regulations affecting catfish, crab, shrimp, and tuna processors. These regulations were issued on June 26, 1974, and are not challenged by petitioners. Rather, petitioners challenge several of the regulations promulgated during phase II of the Agency’s proceedings. These phase II regulations covered nineteen separate subcategories. The subcategories were determined by the species of fish being processed; whether there was mechanization in the processing technique; for Alaska processors, the location of the plant, i. e., whether the plant was located in a “population or processing center”; and in some cases, the production capacity of the plant. At issue are the regulations which apply to Alaskan hand-butchered salmon (Subpart P), Alaskan mechanized salmon (Subpart Q), west coast hañd-butchered salmon (Subpart R), west coast mechanized salmon (Subpart S), Alaskan bottomfish (Subpart T), west coast bottomfish (Subparts U and V), Alaskan scallops (Subpart AC), and herring fillets (Subparts AE and AF). See 40 C.F.R. §§ 408.160 ^408.226, 408.290-408.296, 408.-310-408.326.

The effluent which is the subject of the regulations consists of unused fish residuals. This discharge includes heads, tails, and internal residuals of the processed fish. Substantial quantities of water are used at various stages of the plant operations. This water comes into contact with the fish residuals and contains pollutants when discharged. The regulations prescribe limitations on discharge, and utilize three measures of pollution: five-day biochemical oxygen demand (BOD 5); total suspended solids (TSS); and oil and grease (O & G). 2 The regulations establish daily maximum levels and monthly average levels for each subcategory, and are measured in terms of the amount of pollutant per thousand pounds of fish processed.

The prescribed 1977 BPT for processors not located in Alaska, and Alaska processors located in “population or processing centers,” is the installation of screens to trap the larger fish particles before the effluent is discharged from the plant. 3 Residuals trapped by the screens may be disposed of in ways discussed below. Alaska processors in “remote” locations are subject only to limitations on the size of particles in the effluent, a requirement that can be met by grinding the solids before discharge.

By 1983 the fisheries must comply with more rigorous technology requirements and effluent limitations. For nonremote facilities, the Agency directed that a dissolved air flotation unit be installed at each location and that the end-of-pipe effluent be channeled through this system before it is discharged into the receiving water. These regulations apply to all nonremote subcategories except Subpart V, Conventional Bottomfish. There, the Administrator has prescribed aerated lagoons as the BEA. Remote Alaska fish processors will be required by the 1983 regulations to screen the effluent before discharging it into the receiving waters. 4

*803 Scope of Review

The scope of our review in cases like this is well-settled and need not be restated at length. Our task is to insure that the Agency has accumulated sufficient material upon which to make a reasoned decision, reviewed that material, and promulgated regulations that are the result of reasoned decisionmaking. Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416 , 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971); Marathon Oil Co. v. EPA, 564 F.2d 1253 , 1266 (9th Cir. 1977); American Meat Inst. v. EPA, 526 F.2d 442, 453 (7th Cir. 1975). A comprehensive discussion of a reviewing court’s role in this area is contained in Weyerhaeuser Co. v. Costle, 191 U.S.App.D.C. 309, 323-27 , 590 F.2d 1011, 1024-28 (D.C.Cir.1978) (McGowan, J.).

What might be termed a scope of review issue is raised by petitioner’s substantial reliance on evidence accumulated after promulgation of the final regulations and not a part of the record before the EPA. Whether this court may examine such evidence in reviewing the Agency’s actions is addressed below.

1977 Regulations

The 1977 regulations raise four broad areas of dispute. First, the petitioners question the Agency’s decision to divide some subcategories into remote versus nonremote locations. Second, the petitioners contend that the cost of operating the required technology is wholly disproportionate to the resulting pollution control benefits, and that both costs and benefits were erroneously calculated by the Agency. Third, the. petitioners argue that the Agency’s data collection and analysis are fatally flawed. Finally, petitioners argue that even with the required technology in place, they will be unable to comply with the 1977 effluent limitations.

A. Remote v. Nonremote

A large segment of the fishing industry affected by these regulations is located in Alaska. In determining what is the BPT, the Agency concluded that certain Alaska processing facilities not located in “population or processing centers” should be designated “remote” and subject to less stringent requirements. 5 In the population or processing centers, screening is the designated BPT. According to the EPA, the solids trapped by the screens may be transported over land by trucks either to landfills or to other facilities for further processing, such as conversion to pet food, or barged out to sea and dumped at approved locations. Remote locations are not required to screen. Plants in these locations may simply grind the solids before dumping the effluent into receiving waters.

The Agency’s definition of a “population or processing center,” as distinct from a “remote center,” makes our review difficult because .it is neither a closed-ended definition, that is, one stated in terms relatively incapable of future expansion, nor one which sets forth the characteristics of the definitional class in any functional way. Rather, the definition simply states that certain locations are population or processing centers, and leaves open the possibility that other places may nevertheless possess certain undefined characteristics justifying designation as population or processing centers by some later determination. The definition is contained in the regulations as follows: “Any . . facility [in this subcategory] located in population or processing centers including but not limited to Anchorage, Cordova, Juneau, Ketchikan, Kodiak, and Petersburg shall meet the following limitations: . . . See, e. g., 40 C.F.R. § 408.162 (Alaskan Hand-Butchered Salmon Processing Subcategory).

Petitioners do not contest the Agency’s authority to classify point sources as remote and nonremote locations if the record provides support for such a distinction. They contend, however, that: (a) the record provides no support for the remote/nonremote *804 distinction; 6 (b) even if there is a basis for the distinction, designation of the particular named cities as nonremote is unwarranted; and (c) the definitions of a population center and a processing center are impermissibly vague.

The definitional language, as we have indicated, is not commendable for its precision or for the light it sheds upon the Agency’s criteria for determining what is a nonremote location. In view of the evidence before the Agency in this record, however, we cannot say the distinction is entirely unwarranted or arbitrary, and we sustain it as to the specific locations named as population or processing centers. We find somewhat more support for treating as nonremote cities which are population centers, as distinguished from processing centers. Accordingly, we discuss the two aspects of the definition separately.

There is evidence in the record indicating that construction costs are significantly less for plants located in population centers than in other areas. A study conducted by the National Canners Association shows that the average costs of construction in Anchorage (population of 173,800) and in Juneau (16,600) are 1.5 and 1.6 respectively (1.0 represents 1971 costs in western Washington). Record 7342 (hereinafter R.). For remote Alaska, the cost is about 2.5. R. 6905, 7342. The Agency could reasonably infer from these data that the costs of constructing and operating pollution abatement technology in remote versus nonremote areas would differ. The record also provides some support for the conclusion that plants located in population centers have more dependable access to transportation and landfills. Finally, the Agency suggested in the record that nonremote locations enjoyed economic advantages such as lower rates for transportation, access to population (reflected in relatively lower wages), and more power. R. 14302. Although more detailed analysis would have been preferable, these items provide support for the conclusion that construction, installation, and operation of facilities required to meet the guidelines is more expensive at remote locations than at population centers.

The Agency’s distinction between processing centers, i. e., locations with multiple plants, and remote locations finds somewhat less support in the record, but we are prepared to sustain the Agency’s decision, recognizing that in some later review of this regulatory scheme more specific information may be available either to the Agency or to the industry which might require reconsideration of this distinction. The principal justification for the processing center classification was the finding that there were potential cost savings to plants located in processing centers resulting from the possibility of collective treatment techniques. It is less expensive for plants to share certain disposal costs, such as barging or transportation to landfills or reduction facilities, than if each plant bears them alone. The record indicates that a reduction facility was already in existence in Petersburg, which had the potential for common use, and that four plants in Cordova similarly could employ cooperative barging. By contrast, Agency studies concluded that remote locations would have greater difficulty disposing of waste on lands in part because of undependable land transportation and in part because of geographical conditions making construction and operation of landfills very expensive. We think the assumption that cost savings are available to plants in processing centers is sustainable, at least on the data available when the regulations were initially promulgated. A regulation which recognizes the possibility of pollution treatment through sharing facilities or costs is consistent with the Act.

The current definition of a “population or processing center” is admittedly vague. The regulations as written, how *805 ever, do indicate that population (173,800 for Anchorage and 16,600 for Juneau), plant concentration within the city limits (three phase II plants in Ketchikan and four in Cordova and Petersburg), and relative ease of access to elements which reduce the cost of installing and operating the prescribed technology, are criteria considered by the Agency. We must afford the EPA substantial discretion to implement the mandate of the Act. We would expect that as later category questions arise, the agency will make the definition more precise, and that the more precise definition will have an adequate basis in demonstrated cost savings for waste treatment. We do not, at this point, set aside the Agency’s distinction between nonremote and remote as impermissibly vague.

B. Cost-Benefit Comparisons and the Question of Screening

The parties dispute whether or not the requirement of screening plus barging or land-based disposal was reached after a proper evaluation of costs and benefits. The disagreement extends both to questions of interpreting the statute and to whether the Agency followed the Act even assuming its own interpretation is correct.

Section 304(b)(1)(B) of the Act provides in part:

We think it plain that, as a general rule, the EPA is required to consider the costs and benefits of a proposed technology in its inquiry to determine the BPT. The Agency has broad discretion in weighing these competing factors, however. See Weyerhaeuser, supra, 191 U.S.App.D.C. at 346 , 590 F.2d at 1048 . When considering different levels of technology, it must be shown that increased costs are wholly disproportionate to potential effluent reduction before the Agency is permitted to rely on a cost-benefit comparison to select a lower level of technology as the BPT. This conclusion is consistent with the interpretation of section 304(b)(1)(B) given in the Conference Report on the bill which ultimately became the Act. The Report states:

Congressional Research Service, A Legislative History of the Water Pollution Control Act Amendments of 1972 at 170 (1973) (hereinafter L. H.).

It is relevant in this case, moreover, to consider the definition of the benefits the Agency is directed to weigh. We agree with the Agency’s contention that Congress intended BPT standards to be based primarily on employment of available technology for reducing effluent discharge, and not primarily on demonstrated changes in water quality. See 40 Fed.Reg. at 55771; EPA v. California, 426 U.S. 200, 204-05 , 96 S.Ct. 2022 , 48 L.Ed.2d 578 (1976). Congress was aware that prior enforcement efforts based on water quality standards had not been successful. It determined, accordingly, that the Agency should have the authority to require effluent reduction benefits as defined by the amount or degree of reduction achieved by a level of technology applied to discharge, without the necessity of demonstrating the incremental effect of that technology on the quality of the receiving water. As the D.C. Circuit explained in Weyerhaeuser, supra, 191 U.S.App.D.C. at 339-42 , 590 F.2d at 1041-44 , the “effluent reduction benefits” referred to in the Act are not primarily water quality benefits; rather, “[e]ffluent reduction occurs whenever less effluent is discharged, i. e., whenever a plant treats its wastes before discharge.” Id., 191 U.S.App.D.C. at 342 n. 49, at 1044 n. 49.

*806 In light of these principles, we evaluate the petitioners’ claim that the Agency overestimated the benefits of screening and improperly ignored or underestimated the benefits of grinding. Petitioners assert that grinding should have been considered expressly by the Agency as an alternate technology. It was used by a significant number of plants before the regulations became effective, and the Agency expressly permitted grinding for plants in remote areas. Petitioners assert that the Agency’s failure to consider grinding makes its analysis deficient and invalidates the cost-benefit determination that justifies prescribing screening for nonremote processors. According to petitioners, the record demonstrates that “every ‘adverse environmental impact’ identified by respondent is 'either eliminated by grinding and dispersion or not alleviated by screening at all.” Reply Brief of Petitioners at 11.

We do not understand petitioners to dispute seriously that removing solids by screening is superior to grinding in terms of reducing both the total amount of fish solids and pollutants as defined by the measures used by the EPA. Their point, rather, is that the EPA’s determination of BPT must be evaluated by taking into account the methods of disposing of the collected fish products that are permitted by the regulations in question. Petitioners contend that the ultimate reduction in pollution achieved by screening is the same as grinding, since the EPA permits screened solids to be barged and dumped at specified ocean sites located only xh to 2xh miles from the processing plants. We think petitioners’ reasoning is flawed in various respects.

Barging of solids for dumping at ocean sites is only one of various waste disposal alternatives permitted. Transportation to landfills or pet food processing plants, where the by-products can be sold to the food processors, are also possible methods of handling collected solids. It was reasonable for the Agency to find that given a choice between barging and transporting to a land-based reduction center, some plants would choose the latter. See 40 Fed.Reg. at 55774. To the extent that screened solids are not returned to the receiving waters, fewer pollutants are discharged, and presumably TSS, BOD and O & G are also reduced. We do not think the EPA is required at this stage of the proceedings to demonstrate that the amount of effluent entering the receiving waters would be reduced to the same degree at each site by application of BPT.

The Agency says further that the receiving waters into which it approved discharge of barged solids “are not the same waters” as those used by processors to discharge their ground effluent, because the approved sites are farther offshore. While the petitioners argue that this is irrelevant, there is support for the Agency’s distinction in the legislative history of the Act. 7 The *807 Agency has not explained to this court as clearly as it might have how the asserted water quality benefits of discharging a given amount of effluent farther offshore should be considered within the statutory framework of technology-based, not water quality-based, pollution limitations. But even assuming, contrary to our conclusion in the previous paragraph, that application of the BPT would result in no decrease in the amount of pollutants discharged into receiving waters (both nearshore and offshore), we think that the Act permits the EPA to consider an improvement in near-shore water quality to be an “effluent reduction benefit.” 8 Cf. American Iron & Steel Inst. v. EPA, 568 F.2d 284 , 297 (3d Cir. 1977) (affirming Agency’s cost-benefit analysis, noting that Agency considered the harmful effects of pollutants in question). No doubt Congress sought in the Act to do more than simply have the EPA relocate the same amount of effluent discharge, but we doubt that it prohibited the EPA from requiring relocation of that discharge where environmental benefits of such a relocation have been demonstrated. 9

The record contains evidence that because of inadequate tidal dispersion in various locations affected by the regulations, discharges by processors caused aesthetic and environmental harm by forming sludgebeds at or near beaches and shoreline. The Agency cites studies showing such effects at Kodiak, Cordova, Petersburg, and Anchorage. See R. 11567-11614. The Agency also considered a study by the Canadian Environmental Protection Service which indicated that effluent from fish processing facilities can have harmful environmental effects up to one mile from the discharging facility, altering the environmental balance of the receiving waters and creating toxic waste products. See 40 Fed. Reg. at 55771. The dumping sites approved by the Agency are located in deeper waters and in areas of high tidal activity. These factors facilitate dispersion and decomposition of discharged fish solids.

The petitioners direct our attention to a study of the Petersburg facilities. The document provides qualified support for the conclusion that the nearshore currents at Petersburg are sufficiently strong to prevent sludgebed accumulations. The EPA counters that the study supports a conclusion that grinding and dispersal nearshore has harmful environmental consequences, in particular because it substantially reduces the oxygen in receiving waters. We do not think, therefore, that the study is conclusive of the petitioners’ contentions even as to the Petersburg site, and in any event it does not demonstrate the regulation is inappropriate for the industry as a whole. We cannot find erroneous the Agency’s conclusion regarding inadequate tidal action and environmental effects of discharge into nearshore receiving waters. It was not an abuse of discretion for the Agency to consider an improvement in nearshore water quality as one factor in support of the effluent limitations.

The costs of screening were also considered by the Agency. For example, the Agency projected that for mechanized Alaskan salmon canning plants, “Grinding costs ranged from $30,000 for a 20 ton per-day plant to $54,000 for a 150 ton per-day plant respectively. Screening and barging costs varied from $72,000 to $146,000 for the *808 same plants.” R. 15061-62. The figures for Alaskan Bottom Fish indicate that grinding costs ranged from $20,000 capital outlay and $50/day operation and maintenance (O & M) for a 13.6 ton per day plant to $38,000 capital and $60/day O & M for a 105.6 ton per-day plant. Screening and barging costs varied from $55,000 capital outlay and $150/day 0 & M for the smaller plant to $98,000 capital outlay and $190/day O & M for the larger plant. R. 14859.

After estimating the costs for affected subcategories, the Agency concluded the total internal costs of the 1977 effluent limitations would be $6.2 million for investment and $1.3 million of annual expenditures. 40 Fed.Reg. 55777. External costs included a minor effect on prices and the expected closure of some processing plants because of inability to comply economically with the regulations. See id. The effect of these closures on the domestic industry capacity was anticipated to be small. Id.

Petitioners argue that the number of plants estimated to close as a result of the regulations demonstrates that the costs of implementing the technology are wholly out of proportion to the effluent reduction benefits, and thus that the regulations do not prescribe a practicable technology. We do not find the Agency’s action can be set aside on this issue.

Precisely how many plants in nonremote locations the EPA estimated would close as a result of the 1977 BPT is not completely clear from the record. In the Preamble to the Regulations, 40 Fed.Reg. 55777, the Agency states only that “a number of small plants are projected to be adversely affected by the effluent limitations.” The record shows that in affected subcategories 28 out of 172 plants were projected to close as a result of the 1977 BPT. In its brief, the Agency argues that for several reasons, stated in the record, its original estimate of the number of plant closings was too high. It does not, however, point to any revised estimate. Thus, the most concrete estimate available is that contained in EPA’s Economic Analysis. These data also disclose that in Alaska nonremote subcategories— the Alaska subcategories where screening is the BPT — seven out of sixteen plants were predicted to close as a result of inability to meet BPT. (R. 15055 and 15073). 10

Petitioners agree Congress contemplated that implementing BPT might result in plant closures in some industries. See American Iron & Steel Inst. v. EPA, 526 F.2d 1027, 1052 (3d Cir. 1975). The proportion of plants estimated to close in the nonremote Alaska subcategories (57% for nonremote Alaskan fresh and frozen salmon and 33% for nonremote Alaskan salmon canning) is substantially higher than that approved in some other cases. See, e. g., Weyerhaeuser, supra, 191 U.S.App.D.C. at 345 , 590 F.2d at 1047 (summarizing facts in American Paper Inst. v. Train, 177 U.S.App.D.C. 181, 191-92 , 543 F.2d 328 , 338-39 (D.C.Cir.), ce rt. dismissed, 429 U.S. 967 , 97 S.Ct. 398 , 50 L.Ed.2d 335 (1976)) (out of 270 mills and 120,000 people, 8 mills estimated to close and 1800 people laid off; in three subcategories, of 30 mills, 3 estimated to close). The Agency determined, however, that the effect on prices of implementing the BPT would be small: “price increases generally in the range of 0.3 to 0.5 percent are projected,” 40 Fed.Reg. at 55777. It also found that “domestic industry capacity is not expected to be affected by the potential closure of these particular small plants.” Id. The percentage of estimated *809 plant closures in the seafood processing category generally is low. Given these findings, the estimated number of plant closings in the nonremote Alaska subcategories, standing by itself, does not invalidate its cost/benefit analysis or require us to set aside the Agency’s determination that the required technology was practicable.

The Agency need not balance the costs of compliance against effluent reduction benefits with pinpoint precision, see Weyerhaeuser, supra, 191 U.S.App.D.C. at 146 , 590 F.2d at 1048 ; American Iron & Steel Inst., supra, 568 F.2d at 297, in part because many of the benefits resulting from the effluent reduction are incapable of precise quantification. Cf. American Petroleum Inst. v. EPA, 540 F.2d 1023, 1038 (10th Cir. 1976), cert. denied, 430 U.S. 922 , 97 S.Ct. 1340 , 51 L.Ed.2d 601 (1977) (“The value of the resulting benefits is not capable of present-day determination.”). See also Appalachian Power Co. v. Train, 545 F.2d 1351, 1361 (4th Cir. 1976) (“[W]e [reject] Industry’s contention that benefits derived from a particular level of effluent reduction must be quantified in monetary terms . This reflects the simple fact that such benefits often cannot be reduced to dollars and cents.”).

The Agency, upon consideration of the effluent reduction benefits thought to be achieved by screening, both the water quality benefits and the amount of pollutants discharged into the receiving waters, determined that the costs of screening were justified. We conclude the Agency complied with the Act’s mandate to consider the costs of technology in relation to effluent reduction benefits.

C. Accuracy of the Agency’s Data and Methodology

Petitioners argue that important segments of the EPA’s data collection and the analysis which led to formulation of the effluent limitations were erroneous, with the result that the limitations are arbitrary. The petitioners’ objections to the Agency’s methodology are sufficiently well taken so that the Agency should reconsider such matters at the times specified by the statute for review of the regulations. Given the limitations the Agency faced when it adopted industry standards for the first time, we rule that there was a sufficient basis for promulgating the regulations as an initial matter. Under this standard, despite the petitioners’ objections, we think the 1977 effluent limitations were the product of reasoned decisionmaking, adequately supported by information available to the Agency.

In determining the effluent guidelines for the subcategories, the Agency employed a model plant analysis. Because the Agency determined that treatment practices in the industry were inadequate, it constructed a model plant utilizing processing data gathered from operating processors during the 1973 season, and financial data from 1968 through 1972. The Agency used peak flow in gallons per minute as the major determinant of the size and cost of treatment equipment which would be required.

As we understand petitioners’ challenge, it is that the Agency’s model plant for nonremote Alaskan canned salmon facilities (and presumably other subcategories as well) underestimated the amount of fish processed per hour at actual processors, since the data gathered by the EPA were collected during the 1973 season, an unproductive season not representative of most years. As a result, petitioners say, the model plant analysis underestimated the costs of installing and operating required screening technology. The increased costs would result from the necessity of installing larger screens to handle the greater flow, and possibly also from requirements for more trucks or barges to handle the screened solids.

The Agency’s graph representing nonremote Alaskan canned salmon states that the average processing season is 42 days long. Petitioners claim that the average processing season is actually 14 to 30 days long. The graph shows the average working day during the processing season to be 18 hours. Petitioners claim there is only one crew per *810 plant and that the crews rarely work 18 hours a day. The Agency correlated its figures with the annual production rate and determined that large canners processed approximately 8.3 tons per hour, the medium-sized processors 5.0 tons, and the small plants less than 1.1 tons. Petitioners claim that, in fact, large processors on occasion process up to 40 tons per hour, the medium plants 20 tons, and the small fisheries 10 tons. To make petitioners’ claim more concrete, if a large processor had a water flow rate of 2.2 gallons per minute (the EPA figure) and processed 40 tons of raw material per hour instead of the model plant figure of approximately 8, the cost of compliance with the regulations, while perhaps not being exactly five times as expensive as predicted by the EPA, would nevertheless be substantially higher than the predicted figure.

The Agency defends the model plant as being designed on the basis of figures obtained from actual processors. The Agency also supports its analysis with at least two further arguments. It first notes that the relation, if any, between tons of fish processed per hour and the cost of installing appropriate screening technology is not linear. One cost comparison made by the Agency indicated that a facility which processes ten times as much fish per hour would probably spend only 1.4 times as much as the small facility for the screening technology. R. 7343-7347. Thus, any underestimation regarding plant production would not affect the Agency’s cost estimations to such a degree that it's analysis must be set aside.

The Agency’s other argument is that several of the facilities sampled had or could have had a water use rate of substantially less than 2.2 gallons per pound of raw material, the figure assumed for the model plant. Therefore, even if the tons of raw material processed per hour at a plant were greater than the EPA estimated, its cost figures were not unreasonable estimates, because the plant could process more than 8.3 tons of material per hour using the same flow rate as the model plant, and the cost of installing screens to handle the flow rate would thus be similar to that of the model plant.

The Agency may or may not have constructed the model plant with complete accuracy, but that is not the question for this court. After consulting with members of the processing industry and other knowledgeable sources, the EPA engaged in sampling procedures designed to collect representative data given the constraints imposed by time and difficulties of data collection.

The court’s treatment of a somewhat similar challenge in American Iron & Steel Inst., supra, 568 F.2d at 300-01, is instructive:

The Agency had relevant information before it and considered this information in formulating its production and cost estimates. We decline to second guess the Agency’s expert determinations as to the *811 model plant, since there is adequate support for those conclusions in the record.

The Agency itself recognized that its data collection was not as thorough as it otherwise would have been: “The time constraints imposed by the statutory deadlines precluded the Agency from conducting an exhaustive sampling program.” One of those time constraints was the decision in National Resources Defense Council v. Train, 6 E.R.C. 1033 (D.D.C.1973), rev’d in part and remanded, 166 U.S.App.D.C. 312 , 510 F.2d 692 (D.C.Cir.1975). Similarly, some of petitioners’ critique of the Agency’s methodology is simply an elaboration of shortcomings which the Agency itself considered:

40 Fed.Reg. at 55771.

Petitioners’ insistence that the amount of fish processed per hour must be the major design parameter for determining the cost of complying with the effluent limitations is based in large part on a study conducted by Dr. James Bray, an employee of the University of Washington. Dr. Bray concluded that significantly more fish are processed per hour in real plants than is reflected in the hypothetical processing load of the Agency’s model plant. The study by Dr. Bray is entitled “Review of ‘Economic Analyses’ of Effluent Guidelines, Seafood Processing Industry,” and is dated August 5, 1976. A similar study relied on by petitioners was conducted by Michael D. Swayne and is dated May 17,1976. Finally, a study conducted by Dr. Lloyd Fisher, also a University of Washington professor, entitled “Statistical Review of the Development Document for Effluent Guidelines and New Source Performance Standards for the [Seafood Industry],” was also released after promulgation of the guidelines.

After detailed discussion, all three studies conclude that various statistical and analytical errors by the EPA impugn the validity of the effluent guidelines. Although petitioners raised some questions regarding aspects of EPA’s data collection and analysis during the Agency proceedings, the most telling aspects of petitioners’ critique is based on the analysis contained in the three studies. The analysis was made public after the final regulations were issued and the petition for review was filed, and makes many criticisms which were not brought to the attention of the Agency during the proceedings before promulgation of the regulations.

We evaluate petitioners’ arguments based on these studies in order better to understand the issues before us here. The studies, and arguments based thereon, are helpful in understanding the problem faced by the Agency and the methodology it used to resolve it. To a limited extent, therefore, the post-decision studies can be deemed a clarification or an explanation of the original information before the Agency, and for this purpose it is proper for us to consider them. Bunker Hill Co. v. EPA, 572 F.2d 1286, 1292 (9th Cir. 1977). The comments we have already made are to some extent drawn from this supplementary material. We do not think it is appropriate, however, for either party to use post-decision information as a new rationaliza *812 tion either for sustaining or attacking the Agency’s decision. Id. We think, therefore, it is inappropriate to rely on the specific conclusions of those studies to show that the effluent limitations were not the product of reasoned decisionmaking. See FMC Corp. v. Train, 539 F.2d 973, 978 (4th Cir. 1976); American Iron and Steel Corp. v. EPA, 526 F.2d 1027, 1050, 1057 (3d Cir. 1975), mandate recalled, 560 F.2d 589 (1977), cert. denied, 435 U.S. 914 , 98 S.Ct. 1467 , 55 L.Ed.2d 505 (1978); Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 327 , 486 F.2d 375, 394 (D.C. Cir.1973), cert. denied, 417 U.S. 921 , 94 S.Ct. 2628 , 41 L.Ed.2d 226 (1974). As the court in American Iron and Steel Inst., supra, said, “We believe that it would be highly disruptive to the administrative process to allow a company to sit back and wait until the regulations were published in final form before coming forth and contending that the agency had failed to consider [certain facts].” 526 F.2d at 1050.

If the studies showed that the Agency proceeded upon assumptions that were entirely fictional or utterly without scientific support, then post-decisional data might be utilized by the party challenging the regulation. Cf. American Petroleum Inst. v. EPA, supra, 540 F.2d at 1034 . That, however, is not the case here. Viewed as material that illuminates the original decision made by the Agency, the studies in question do not undermine our conclusion that the effluent limitations were the products of reasoned decisionmaking.

The Act provides for annual revision of guidelines for effluent limitations-— such as the challenged regulations — promulgated under section 304, 33 U.S.C. § 1314 . Section 304(b), 33 U.S.C. § 1314 (b). The Act also establishes variance procedures, section 301(c), 33 U.S.C. § 1311 (c). The standards promulgated by the Agency governing its exercise of discretion in granting variances to affected processors are similar to those which have been upheld for other industrial categories. See 40 C.F.R. § 408 .-162; Weyerhaeuser, supra, 191 U.S.App.D.C. at 336 , 590 F.2d at 1038 . Moreover, the Act requires review of 1983 regulations every five years after their promulgation, section 301(d), 33 U.S.C. § 1311 (d). At these points the Agency will be required to justify its regulations in terms of the more extensive data developed since the regulations were first promulgated. In an appropriate case, moreover, a petition for reconsideration may be filed with the EPA to consider whether evidence such as that offered by petitioners requires the Agency to review its original actions. Oljato Chapter of the Navajo Tribe v. Train, 169 U.S.App.D.C. 195 , 515 F.2d 654 (D.C.Cir.1975). Thus, there are mechanisms for the Agency to consider evidence developed after promulgation of the 1977 regulations.

D. Pollutant Variability

Petitioners challenge further the Agency’s findings as to the causes of variability in the pollutant load. Petitioners claim that even if they install a type of screen recommended by the Agency, many facilities will, at certain times, be unable to meet the effluent limitations specified in the regulations, because the amount of pollutants emitted by affected facilities is subject to extreme variability from causes over which the processors have no practical control. As a result, the petitioners contend, the Agency did not sufficiently demonstrate that the technology can be applied practically to comply with the effluent limitations. A related contention is that the Agency relied improperly on “in-plant process changes” in determining the effluent guidelines.

To establish the effluent limitations, the EPA sampled outflows in each subcategory. The measurement of TSS and O & G fluctuated substantially, both among plants in the same subcategory and at various single plants at different times. In the nonremote Alaskan mechanized salmon subcategory, for instance, the TSS measurement varied from 13.2 to 44 per thousand pounds of raw material, and the O & G measurement varied from 3.19 to 29 lbs. per thousand. The daily and monthly maximum limitations established by the Agency are based on the average of these highly variable measurements.

*813 The Agency found that the primary cause of variability, in single plants and also among plants in the same subcategory, is disparate water and waste management practices. Petitioners contend that the primary causes of pollution fluctuation are other factors, most of which are uncontrollable, including the species and sexual maturity of the fish being processed, the degree of mechanization in each processing plant, and the condition of fish on delivery to the plant. While the first two factors cited by petitioners do bear upon pollution variability, we cannot say that the Agency was incorrect in concluding that they were insignificant in comparison to fluctuations caused by water and waste management practices. Petitioners focus on the salmon sulpcategories. Despite petitioners’ claim that the species of fish affects the waste-load, the EPA found that on the average, each species of salmon produced the same amount of waste, roughly, one-third of the whole fish. R. 11573, 11587. A study by the National Canners Association also supports this figure.

Similarly, the petitioners’ challenge to the Agency’s test for distinguishing between mechanized and nonmechanized plants is not persuasive. The Agency’s basis for distinguishing mechanized from hand-butchered processors is whether the plant uses an “iron chink” machine to butcher the whole fish. Both parties agree that use of the machine causes a significant increase in waste material. Petitioners contend, however, that other machines also contribute to the wasteload, and that the Agency erred in not taking these machines into account in establishing the mechanized subcategory. Petitioners point particularly to the use of can-filling machines in some plants. The petitioners do not, however, cite any study demonstrating that mechanical fillers substantially affect wasteload. Despite the fact that some waste is produced during the canning process, there is no evidence in the record that this factor would account for the variability in the pollution figures. On the contrary, the EPA analyzed the sources of the wasteload in the salmon canning industry and failed to find a significant increase in TSS resulting from the canning procedure. R. 14619-20. The apparent reason is that even though the mechanized fillers spill more fish meat than hand packing operations, the spilled meat is recaptured at patching tables where light cans are hand filled with extra meat. R. 14756.

The Agency agreed with petitioners that the sexual maturity of the fish and the condition of the raw material on delivery to the processing facility (fresh vs. old and hard vs. soft) are uncontrollable factors which can affect plant raw waste loads. 40 Fed.Reg. at 55773. It contends, however, that by sampling effluent at specific plants over several weeks, it was able to take into account the variability in a plant’s statistics caused by this factor. In other words, the Agency has determined that the effluent guidelines can be met, even though a plant’s raw waste load may vary from day to day because of the condition of the fish on delivery to the plant. Within the limitations of the Agency’s data gathering procedures — limitations which we have already held do not require setting aside the guidelines — the Agency’s data collection would appear to have taken into account variability caused by the condition of the product upon delivery to the plant. The guidelines contain, moreover, both daily and monthly limitations, and the daily limitation permits unusually high discharges of effluent as long as the monthly limit is met.

As noted, the Agency determined that the major cause in variable pollution loads was the difference in waste and water management practices at single plants and between different plants in the same subcategory. The Agency focused on water use as the major treatment procedure, in addition to screening, which would enable plants to meet the guidelines. It concluded that “The seafood processing industry in general, and particularly the Alaskan processing facilities, ignores all concepts of water conservation.” In the course of collecting data, the EPA observed the following water management practices:

40 Fed.Reg. at 55773. Petitioners agree with respondent that “there is extensive water use” by some processors, and agree that “excessive water use should be reduced.” Brief of Petitioners at 35. Their contention, discussed later, is that turning off the water at the times suggested by the EPA and employing spring-loaded hose nozzles, 40 Fed.Reg. at 55773, will not assist processors in meeting the guidelines. For now, however, the important point is that petitioners have not contended that many of the water use practices recommended by the Agency are either impracticable or extremely costly.

Our first conclusion is that it is permissible for the Agency to base its cost calculations and effluent guidelines on the assumption that processors will use 2.2 gallons of water per pound of raw material. The figure was the average flow rate at plants sampled by the Agency and by the National Canners Association in an earlier study relied on by the Agency. See American Iron and Steel Inst., supra, 568 F.2d at 300, 305-06 (noted by the court in Weyerhaeuser, which dealt with a similar water use challenge). The Agency does not require operators to use a specific quantity in gallons per pound of raw material. See Weyerhaeuser, supra, 191 U.S.App.D.C. at 357 , 590 F.2d at 1059 . Processors may, if they wish, install somewhat more extensive and expensive screens and use a flow rate greater than the Agency norm. We find, as did the court in Weyerhaeuser, that the EPA could determine that processors could practicably reduce their water use to a given gallon amount per pound of raw material. In determining the effluent guidelines and the cost of complying with those guidelines, the Agency could assume that plants in affected subcategories would use the flow rate ascribed to the model plant, despite some observed variability in the amount of water used by different processors.

This problem differs from that addressed by the court in Marathon Oil Corp. v. EPA, 564 F.2d 1253 (9th Cir. 1977). There, in a case involving a challenge to the terms contained in a discharge permit, the Agency imposed an absolute limit on the number of pounds of pollutant that could be discharged from a drilling rig per day, despite the large variance in the amount of effluent flow from the decks in question. The court set aside the single-number limit. 564 F.2d at 1269-70 . In our case the Agency has not regulated permissible discharge by setting a pound limitation that is independent of the production rate. The limitation is, instead, proportional to the production rate and, as it was based on adequate support, is not an arbitrary limit.

The variability was much greater in FMC Corp. v. Train, 539 F.2d 973 (4th Cir. 1976), than in the case at bar. Whereas the gallons per pound of product ranged from .3 to 5 in one synthetic subcategory, id. at 980 , the largest range in the EPA’s data for the salmon subcategory was 2.19 to 2.45. See Weyerhaeuser, supra, 191 U.S.App.D.C. at 358 n.87, 590 F.2d at 1060 n.87 (distinguishing FMC on similar grounds). Although the National Canners Association study found the gallons per pound of raw material to be as low as .58, this variability would tend to favor the processors since the regulations were based on the higher figures collected by the EPA.

Further, it does not appear from the FMC opinion that the Agency attempted to provide any reason for the extreme variability. In the present case, however, the EPA has provided a reasonable explanation for the variability in pollutant and water use measurements: the cheapness of using even large amount of water and inattention *815 of the industry to water management practices. For these reasons, this case is not inconsistent with the court’s ruling in FMC regarding statistical variability.

The Agency did not appear to make extensive findings regarding the extent to which individual plants in a subcategory could reduce their pollutant loads by reducing water use in the ways suggested in the Federal Register. It did, however, determine that by employing the suggested water practices where appropriate, water/solids contact, and thus pollutant load, could be reduced substantially. See, e. g., R. 14749-50. In addition, data collected at a mechanized salmon plant indicated that water use during cleanup operations constituted about 20% of the plant flow and 10% of the total BOD and TSS. A similar study at a hand-butchered salmon cannery indicated that water use during cleanup was about one-third of the total flow, and was responsible for about one-third of the BOD 5 and TSS measurements. R. 14619-20. Petitioners state that “The water reduction outlined [by the Administrator] will not affect effluent loads — and it is effluent load which affects the ability to meet the effluent guideline numbers.” Brief of Petitioners at 36. The reason, contend petitioners, is that the recommended water reductions occur at times when the water does not come into contact with the fish solids, and thus no additional pollutants are created by the excessive use. We find, contrary to the petitioners’ argument, sufficient evidence in the record to support the Agency’s findings that water control will substantially reduce effluent load.

More serious are the petitioners’ questions regarding the waste management practices suggested by the Agency. The Agency referred, for instance, to “dry cleanup methods such as shoveling solid waste into bins prior to water cleanup” and using conveyor belts instead of water to transport the waste. The Agency has noted that processors can effect substantial pollution reduction by the adoption of better waste management practices. Petitioners contend such a change would require substantial redesign of many processors’ operating facilities at a great cost, and that therefore these recommendations constitute “in-plant process changes” which are beyond the authority of the Agency to enforce, absent a finding that they are normal practice within the industry. Weyerhaeuser, supra, 191 U.S.App.D.C. at 357 , 590 F.2d at 1059 ; FMC, supra. The Agency did state that “good housekeeping” measures were considered normal practice within the industry, 40 Fed.Reg. at 55773. But the Agency also found that water and waste management practices in the seafood industry were generally inadequate, and found that “none of the plants” in the mechanized salmon subcategories employed any waste management practices. While a finding of general inadequacy is not necessarily inconsistent with the conclusion that good housekeeping measures were considered normal practice within the industry, neither are the two statements easily reconciled.

We understand the Agency’s position to be that such waste management practices are not an element of BPT and are not required in order to meet effluent limitations. We agree with this conclusion. The model plant, which formed the basis for the limitations set by the Agency, did not employ such waste management practices. Were the Agency to rest its regulations on the premise that BPT required extensive waste management changes, we would face the question whether or not such technologies constitute in-plant process changes that are beyond the authority of the Agency to require for its 1977 limits. If we faced that issue, the Agency might successfully contend that waste management is simply an aspect of a treatment of end-of-pipe discharge and not a part of the manufacturing process. The argument would be that inasmuch as fish Waste products must be discarded and are directly in the channel of the effluent discharge, its disposal is properly viewed as effluent treatment apart from the existing manufacturing process. See American Iron & Steel Inst., supra, 568 *816 F.2d at 306. 11 We need not, however, reach that question since the effluent limitations can be supported by the Agency’s findings and conclusions relating to the flow rates at the model plant.

As with other aspects of petitioners’ challenge, it seems to us that the variance procedures and periodic statutory review mechanisms are adequate vehicles to correct whatever errors the Agency made in its initial, admittedly not completely thorough, effort to formulate effluent guidelines. The Agency must, in good faith, be receptive to whatever new evidence petitioners may bring forth regarding the accuracy of the Agency’s initial regulations, and to adjust its guidelines if the evidence shows petitioners’ inability to comply with them.

1983 Regulations

The regulations promulgated by the Administrator for 1983 are to reflect the “best available technology economically achievable.” Section 301(b)(2)(A), 33 U.S.C. § 1311 (bX2)(A). For the regulations to be affirmed, the Agency must demonstrate that the technology required is “available” and the effluent limitations are “economically achievable.”

A. Dissolved Air Flotation Unit.

A dissolved air flotation unit is the prescribed technology for the following subcategories: west coast hand-butchered salmon (Subpart R), west coast mechanized salmon (Subpart S), non-Alaskan mechanized bottomfish (Subpart V), and non-Alaskan herring fillet (Subpart AF). 12

The system operates as follows: Effluent passes into a holding tank. Air enters the effluent under pressure and attaches to solid particles. Buoyed by the air, the particles rise to the surface and are skimmed off. The particles can be used as animal feed or fertilizer. The clarified water is withdrawn from the bottom of the tank. Before entering the receiving waters, this effluent remainder must be within certain maximum limitations. The regulations set forth limitations for BOD 5, TSS, O & G, and acidity for each category.

The numerical limitations were based on the assumption that a DAF unit will reduce BOD b by 75%, TSS by 90%, and O & G by 90%. Petitioners claim these figures were the result of a single study conducted by the British Columbia seafood industry in conjunction with the Canadian Fisheries Research Board. The study measured the amount of reduced pollutants in the effluent from a salmon processing plant. The DAF unit reduced the BODb by 80%, the TSS by 90%, and the oil and grease by 95%. Petitioners argue that the Agency was arbitrary and capricious in basing its 1983 technology regulations on a single study.

The legislative history of the 1983 regulations indicates that regulations establishing BEA can be based on statistics from a single plant. The House Report states:

L.H. at 798; FMC, supra, 539 F.2d at 983-84 ; American Iron & Steel Inst., supra, 526 F.2d at 1058.

Although only one salmon study was in the record, the Administrator also considered numerous other DAF studies involving herring, groundfish, stickwater, sardines, shrimp, tuna, mackerel, scabbard, yel *817 low croaker, and menhaden bailwater. Each of these studies revealed a substantial reduction in pollution levels. Although the pollution reduction in some studies was not as dramatic as others, the EPA is not charged with burden of showing that all DAF units could meet the limitations, but rather that the best existing DAF units can meet the limitations. Further, to the extent that some of these studies are best viewed as implicating “transfer technology,” the Agency did not misuse its discretion in finding the technology to be transferable to the subeategories at issue. See generally Weyerhaeuser, supra, 191 U.S.App.D.C. at 351 n. 70, 590 F.2d at 1054 n. 70; C & H Sugar Co. v. EPA, 553 F.2d 280, 286 (2d Cir. 1977). The EPA’s data base was sufficient to show that the technology required to meet the 1983 limitations is “available.” See American Iron and Steel Inst., supra, 526 F.2d at 1062.

As with the 1977 limitations, the petitioners have submitted various studies that were not made part of the administrative record and were conducted after the final regulations were promulgated. These studies include new data allegedly demonstrating that the DAF units will not achieve the 1983 limitations. As discussed above, this court will not consider evidence for the first time on review that should properly have been submitted to the Agency for its consideration. There are ways, described earlier in this opinion, for petitioners to bring these studies to the Administrator’s attention.

The next question is whether the Agency properly evaluated the costs of meeting the 1983 guidelines. In describing the role of costs in promulgating BEA, the Conference Report stated:

L.H. at 170. Although the wording of the statute clearly states that the 1983 limitations must be “economically achievable,” there is some disagreement among the circuits as to whether the costs of compliance should be considered in a review of the 1983 limitations. Compare Appalachian Power Co. v. Train, 545 F.2d 1351, 1361 (4th Cir. 1976) with American Iron and Steel Inst., supra, 526 F.2d at 1052. We hold, in agreement with the court in Weyerhaeuser, supra, 191 U.S.App.D.C. at 342-43 , 590 F.2d at 1044-45 , that the EPA must consider the economic consequences of the 1983 regulations, along with the other factors mentioned in section 304(b)(2)(B), 33 U.S.C. § 1314 (b)(2)(B). See also Appalachian Power Co. v. Train, supra, 545 F.2d at 1361 ; American Iron & Steel Inst., supra, 526 F.2d at 1051-52.

Petitioners maintain that the Agency must balance the ecological benefits against the associated costs in determining whether the technology is economically achievable. They cite in support of this proposition Appalachian Power Co. v. Train, supra. In remanding the 1983 regulations affecting electrical power companies, the court there stated:

Id. at 1364-65 (footnote omitted). According to petitioners, the Agency did not consider the incremental benefit to the environment to be achieved by the dissolved air flotation units, and the regulations must be set aside. We cannot agree. As noted by the court in Weyerhaeuser, supra, 191 U.S. *818 App.D.C. at 342-43, 590 F.2d at 1045-46 , the language of the statute indicates that the EPA’s consideration of costs in determining BPT and BEA was to be different. In prescribing the appropriate 1977 technology, the Agency was to “include consideration of the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application.” Section 304(b)(1)(B) (emphasis added). In determining the 1983 control technology, however, the EPA must “take into account ... the cost of achieving such effluent reduction,” along with various other factors. Section 304(b)(2)(B). The conspicuous absence of the comparative language contained in section 304(b)(1)(B) leads us to the conclusion that Congress did not intend the Agency or this court to engage in marginal cost-benefit comparisons. See, e. g., American Iron & Steel Inst., supra, 526 F.2d at 1051.

The intent of Congress is stated in 33 U.S.C. § 1251 (a)(1): “[I]t is the national goal that the discharge of pollutants into the navigable waters be eliminated by 1985 . .” The regulations that will be applied in 1983 are intended to result “in reasonable further progress toward the national goal of eliminating the discharge of all pollutants . . . .” 33 U.S.C. § 1311 (bX2)(A). These express declarations of congressional intent cannot be ignored in determining the reasonableness of the 1983 regulations. So long as the required technology reduces the discharge of pollutants, our inquiry will be limited to whether the Agency considered the cost of technology, along with the other statutory factors, and whether its conclusion is reasonable. Of course, at some point extremely costly more refined treatment will have a de minimis effect on the receiving waters. Cf. Appalachian Power Co., supra, 545 F.2d at 1364, 1365 . But that point has not been reached in these BEA regulations.

The record discloses that the Agency studied the cost of complying with the 1983 regulations. R. 14809. It set forth the cost of compliance for plants that produced various amounts of effluent per minute, both in terms of capital costs and operation and maintenance costs. R. 14810. The projections include estimates of the costs of construction, labor, power, chemicals and fuel. R. 14806. Although land acquisition costs were not considered, the amount of land necessary for the air flotation unit is minimal. In contrast to our conclusion regarding aerated lagoons, the Agency was not arbitrary in concluding that the DAF unit could be installed on existing plant locations without necessitating additional land acquisitions. See R. 15039.

Finally, it does not appear that the cost of complying with the 1983 regulations is unreasonable. The cost of compliance for the Northwest Canned Salmon subcategory, for example, is estimated to be $157,000 for initial investment and $32,000 of annual expenditures for the average size plants. R. 14820. According to the EPA’s economic analysis, the total annual costs of pollution abatement averaged between one and two percent of the total sales figures of each subcategory. R. 15074-75. 13 Depreciation is available for the capital outlays and tax deductions are available for business expenses. The Agency concluded that the benefits justified the costs, and petitioners have not shown that conclusion to be arbitrary or capricious. See American Iron & Steel Inst., supra, 526 F.2d at 1052-53.

Although the number of plants estimated to close as a result of the 1983 regulations was not stated clearly to us, it appears to be a lesser proportion of affected plants than that which we approved for the 1977 regulations. See R. 15072-77. Since Congress contemplated the closure of some marginal plants, we do not consider the regulations to be arbitrary and capricious.

For these reasons, we conclude that the 1983 regulations requiring dissolved air flotation units for West Coast fish processors should be upheld.

*819 B. Aerated Lagoons

The aerated lagoon is the required technology for the non-Alaskan conventional bottomfish subcategory. 40 C.F.R. § 408 .-213. 14 Aerated lagoons are still ponds in which waste water is treated biologically. They are usually three to four feet deep. With oxidation taking place in the upper eighteen inches, the water will remain in the lagoon from three to fifty days. Mechanically aerated lagoons are between six and twenty feet deep and receive oxygen from a floating aerator. R. 14796. Because of the length of detention time, the lagoons must be large in relation to the square footage of a processing plant to handle peak wasteload production.

Petitioners contend that the data are insufficient to determine if the limitations are achievable. The Agency relied on one study conducted with perch and smelt wastewater. R. 3043 et seq. Although the record also makes a vague reference to a shrimp processing lagoon in Florida, the record contains no information about its similarities to or differences from the technology at issue.- The record does not disclose the analytic approach utilized, the transferability of the technology, or even the person or persons who conducted the study. Therefore, we limit our discussion to the smelt and perch wastewater study.

Although the 1983 regulations can be based on information from a single model plant, see discussion supra, the study must demonstrate the effectiveness of the required technology. The smelt perch study does not reach this standard. The study measured the BOD 5 and TSS contained in the wastewater before entry to the lagoon 15 (BOD 5 4.5, TSS 2.3), but the record reveals only the reduction in BOD 5 in the effluent leaving the lagoon. R. 3044. Although the BOD 5 level would comply with the 1983 limitations, there is no indication as to whether the TSS and O & G levels would be sufficiently reduced.

The Agency is charged with the duty of articulating the reasons for its determination. See Appalachian Power Co. v. EPA, 477 F.2d 495 , 507 (4th Cir. 1973). Since there are no data upon which to determine that the TSS and O & G levels will be reduced sufficiently to comply with the 1983 limitations, we remand these regulations to the Agency for further findings.

The final question presented is whether the Agency gave adequate consideration to costs in the determination that aerated lagoons are achievable technology. The major disadvantage of aerated lagoons is that they require large amounts of readily-accessible land. R. 14796. The Agency did not consider the cost of acquiring land in determining the economic impact of the regulations, R. 15039, 15043. The reason given for this omission is that the costs of acquiring land are “site-specific” and vary depending upon the location and surrounding area. Although this may be true, it does not follow that the cost of acquiring land should be completely ignored in determining whether the technology is achievable. Where a significant amount of land proximate to a plant is an inherent requirement of a control technology, the Agency must attempt to determine the economic impact of acquiring the land. The Agency may set forth the amount of land necessary for various size plants, the average cost of *820 land in the vicinity of identified processing plants, and, finally, whether it is reasonable to conclude that land will be available for the aerated lagoons.

We recognize this holding may be at odds with American Iron and Steel Inst. v. EPA, 526 F.2d 1027, 1053 (3d Cir. 1976). The court there determined that the cost of land acquisition should not be considered because it is “inherently site-specific” and that petitioners have the burden of showing “the magnitude of these excluded costs factors.” Id. at 1053 . We respectfully disagree on both counts.

The Agency has successfully projected other average costs that might be termed “site-specific” such as the cost of in-plant process changes, the cost of barging and screening and the cost of other control technology. Each of these costs is subject to variability, depending on plant design and location. R. 15210. Similarly, when a significant amount of land is required for implementation of the regulation, we think the Agency must take land availability or land cost into account in some manner before it can make a reasoned determination that the regulations are economically achievable, especially when actual plant sites can be studied for this purpose. Failure to do so results in an incomplete consideration of the economic impact of the regulations. This is especially true where, as here, it appears many plant sites are located on waterfront areas where proximate land is not available or is in all probability expensive. That the cost or practicality of acquiring land is difficult to discern does not excuse the Agency from making some estimate of those factors. The amount of land necessary for this control technology is quite significant, and we think it is essential that the Agency give express consideration to that aspect of the 1983 regulations.

The burden of proving the measure of the excluded costs should not be placed upon the industry. It is the Agency’s duty to “fully explicate its course of inquiry, its analysis, and its reasoning.” Tanner's Council of America, Inc. v. Train, 540 F.2d 1188, 1191 (4th Cir. 1976). The record clearly states that aerated lagoons will require large amounts of land. R. 14796. The agency must show that the regulations are achievable. These regulations, based on data that fail to consider land acquisition cost or availability, are not the result of reasoned decisionmaking.

We therefore remand the regulation requiring aerated lagoons to the Agency in order that it may promulgate new 1983 regulations for the non-Alaskan conventional bottomfish subcategory. The remaining regulations are sustained upon the reasoning set forth above.

AFFIRMED in part and REMANDED in part.

. The Agency has adopted a multiple-phase approach to promulgating regulations for an industry in other cases, see, e. g., Weyerhaeuser v. Costle, 191 U.S.App.D.C. 309, 318-19 , 590 F.2d 1011, 1020-21 (D.C. Cir.1978).

. BOD 5, five-day biochemical oxygen demand, is a measure of the oxygen-consuming potentialities of organic matter in the effluent. The BOD 5 test measures the extent to which the biological degradation of organic waste matter over a five-day period removes oxygen from the water. This test does not identify what kind of organic material is present in the effluent, but merely indicates the presence of biodegradable matter.

TSS describes the quantity of undissolved solid matter suspended in the effluent. As with BOD, the TSS measure does not identify the nature or the physical and chemical characteristics of the matter present in the effluent.

Oil and grease describes the volume of naturally-occurring fish oil in the effluent.

. Although the Agency prescribed screening as the BPT, “processors may select alternative methods ... to meet the published effluent limitations.” 40 Fed.Reg. at 55774.

. With regard to both the 1977 and 1983 guidelines, the Agency suggested that certain in-plant process changes would be made at affected plants. We discuss this aspect of the case below.

. The remote/nonremote distinction was first developed during hearings on phase I of the EPA’s development of regulations for the canned and preserved seafood processing industry.

. Petitioners do not contest the Agency’s determination that grinding is appropriate for remote Alaska locations. Their dispute is with the Administrator’s conclusion that facilities in nonremote locations must meet more stringent requirements.

. The following exchange occurred between Senator Stevens of Alaska and Senator Muskie:

L.H. 1347^18.

. Congress was clearly concerned in the Act with the failure of prior water quality based pollution statutes.. See Weyerhaeuser, supra, 191 U.S.App.D.C. at 338—42, 590 F.2d at 1040-44 . The shift in the Act to technology based pollution limitation was an effort to free the EPA from the necessity of proving in every case that application of an effluent limitation at a specific site will improve water quality at that site. Id. We doubt, however, whether Congress meant to prohibit the EPA from considering the subcategory-wide water quality impacts.

. We decline at this time to discuss the extent to which, if the EPA relies on water quality evidence in measuring the benefits of requiring a particular technology for a category or subcategory of point sources, it must also consider water quality evidence at particular sites in passing on applications for variances.

. The figures for these subcategories are:

Segment/size (tons per day) No. of plants_ EPA Estimated closures attributed to BPT EPÁ Estimated closures attributed to BPT and BAT

Alaskan Fresh and Frozen Salmon (non-remote)

10 TPD 3 3 3

10-20-TPD 2 1 1

20 TPD 2 None None

Segment/size (tons per day) No. of plants EPA Estimated closures attributed to BPT EPA Estimated closures attributed to BPT and BAT

Alaskan Salmon Canning (non-remote)

50 TPD 8-3 8

50 TPD 1 None None

. We also need not reach the question whether pollution reduction techniques characterizable as in-plant process changes may be required by the Agency upon a finding that they are practicable, or relatively inexpensive, even if they are not now considered normal industry practice.

. The regulations regarding Alaskan mechanized salmon and herring fillets were withdrawn by the Agency and are not at issue in this case.

. The only subcategories' that exceeded this amount were the Alaskan Fisheries. Since the 1983 regulations affecting these subcategories have been withdrawn, their costs were not considered in determining whether the 1983 regulations were achievable.

. The subcategory averages are as follows:

Effluent Limitations

Effluent characteristic Maximum for any i day Average of daily value for 30 consecutive days shall not exceed (pounds per 1,000 lb. of seafood)

BOD 1.2 0.71

TSS 1.5 0.73

Oil and grease 0.077 0.042

pH Within the range 6.0 to 9.0

. BOD 4.5, TSS 2.3.